Loads & Install Packages
if (!require("nnet")) install.packages("nnet")
## Caricamento del pacchetto richiesto: nnet
## Warning: il pacchetto 'nnet' è stato creato con R versione 4.2.3
if (!require("MASS")) install.packages("MASS")
## Caricamento del pacchetto richiesto: MASS
## Warning: il pacchetto 'MASS' è stato creato con R versione 4.2.3
if (!require("e1071")) install.packages("e1071")
## Caricamento del pacchetto richiesto: e1071
## Warning: il pacchetto 'e1071' è stato creato con R versione 4.2.3
if (!require("class")) install.packages("class")
## Caricamento del pacchetto richiesto: class
if (!require("leaps")) install.packages("leaps")
## Caricamento del pacchetto richiesto: leaps
## Warning: il pacchetto 'leaps' è stato creato con R versione 4.2.3
if (!require("glmnet")) install.packages("glmnet")
## Caricamento del pacchetto richiesto: glmnet
## Warning: il pacchetto 'glmnet' è stato creato con R versione 4.2.3
## Caricamento del pacchetto richiesto: Matrix
## Warning: il pacchetto 'Matrix' è stato creato con R versione 4.2.3
## Loaded glmnet 4.1-8
#if (!require("plm")) install.packages("plm")
if (!require("summarytools")) install.packages("summarytools")
## Caricamento del pacchetto richiesto: summarytools
## Warning: il pacchetto 'summarytools' è stato creato con R versione 4.2.3
if (!require("dplyr")) install.packages("dplyr")
## Caricamento del pacchetto richiesto: dplyr
## Warning: il pacchetto 'dplyr' è stato creato con R versione 4.2.3
##
## Caricamento pacchetto: 'dplyr'
## Il seguente oggetto è mascherato da 'package:MASS':
##
## select
## I seguenti oggetti sono mascherati da 'package:stats':
##
## filter, lag
## I seguenti oggetti sono mascherati da 'package:base':
##
## intersect, setdiff, setequal, union
if (!require("ggplot2")) install.packages("ggplot2")
## Caricamento del pacchetto richiesto: ggplot2
## Warning: il pacchetto 'ggplot2' è stato creato con R versione 4.2.3
if (!require("tidyverse")) install.packages("tidyverse")
## Caricamento del pacchetto richiesto: tidyverse
## Warning: il pacchetto 'tidyverse' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'tibble' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'tidyr' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'readr' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'purrr' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'stringr' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'forcats' è stato creato con R versione 4.2.3
## Warning: il pacchetto 'lubridate' è stato creato con R versione 4.2.3
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ lubridate 1.9.3 ✔ tibble 3.2.1
## ✔ purrr 1.0.2 ✔ tidyr 1.3.0
## ✔ readr 2.1.4
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ tidyr::expand() masks Matrix::expand()
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ✖ tidyr::pack() masks Matrix::pack()
## ✖ dplyr::select() masks MASS::select()
## ✖ tidyr::unpack() masks Matrix::unpack()
## ✖ tibble::view() masks summarytools::view()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
if (!require("lubridate")) install.packages("lubridate")
if (!require("mapview")) install.packages("mapview")
## Caricamento del pacchetto richiesto: mapview
## Warning: il pacchetto 'mapview' è stato creato con R versione 4.2.3
if (!require("sf")) install.packages("sf")
## Caricamento del pacchetto richiesto: sf
## Warning: il pacchetto 'sf' è stato creato con R versione 4.2.3
## Linking to GEOS 3.9.3, GDAL 3.5.2, PROJ 8.2.1; sf_use_s2() is TRUE
if (!require("geojsonio")) install.packages("geojsonio")
## Caricamento del pacchetto richiesto: geojsonio
## Warning: il pacchetto 'geojsonio' è stato creato con R versione 4.2.3
## Registered S3 method overwritten by 'geojsonsf':
## method from
## print.geojson geojson
##
## Caricamento pacchetto: 'geojsonio'
##
## Il seguente oggetto è mascherato da 'package:base':
##
## pretty
if (!require("leaflet")) install.packages("leaflet")
## Caricamento del pacchetto richiesto: leaflet
## Warning: il pacchetto 'leaflet' è stato creato con R versione 4.2.3
if (!require("broom")) install.packages("broom")
## Caricamento del pacchetto richiesto: broom
## Warning: il pacchetto 'broom' è stato creato con R versione 4.2.3
if (!require("plotly")) install.packages("plotly")
## Caricamento del pacchetto richiesto: plotly
## Warning: il pacchetto 'plotly' è stato creato con R versione 4.2.3
##
## Caricamento pacchetto: 'plotly'
##
## Il seguente oggetto è mascherato da 'package:ggplot2':
##
## last_plot
##
## Il seguente oggetto è mascherato da 'package:MASS':
##
## select
##
## Il seguente oggetto è mascherato da 'package:stats':
##
## filter
##
## Il seguente oggetto è mascherato da 'package:graphics':
##
## layout
library(nnet)
library(MASS)
library(e1071)
library(class)
library(leaps)
library(glmnet)
#library(plm)
library(summarytools)
library(dplyr)
library(ggplot2)
library(tidyverse)
library(lubridate)
library(mapview)
library(sf)
library(geojsonio)
library(leaflet)
library(broom)
library(plotly)
Data Exlporation and Cleaning
The first step is always to read the dataset and plot the first 5 observations
fire_data <- read.csv("datasets/Fire_Incident_Dispatch_Data_last_50k.csv")
head(fire_data)
## STARFIRE_INCIDENT_ID INCIDENT_DATETIME ALARM_BOX_BOROUGH
## 1 230905-B1937-001-0567 09/05/2023 02:19:04 PM BROOKLYN
## 2 230905-B3923-002-0568 09/05/2023 02:19:36 PM BROOKLYN
## 3 230905-X8897-003-0480 09/05/2023 02:19:43 PM BRONX
## 4 230905-X3466-001-0481 09/05/2023 02:21:00 PM BRONX
## 5 230905-B2448-001-0570 09/05/2023 02:21:26 PM BROOKLYN
## 6 230905-B2448-002-0571 09/05/2023 02:22:35 PM BROOKLYN
## ALARM_BOX_NUMBER ALARM_BOX_LOCATION INCIDENT_BOROUGH
## 1 1937 AUTUMN AVE & FULTON ST BROOKLYN
## 2 3923 N/S EASTERN PWAY & UTICA AVE BROOKLYN
## 3 8897 CROSS BX EXPY- DEEGAN EX TO JEROME AV BRONX
## 4 3466 ADEE AVE & BX PARK E BRONX
## 5 2448 GLENWOOD RD & BEDFORD AVE BROOKLYN
## 6 2448 GLENWOOD RD & BEDFORD AVE BROOKLYN
## ZIPCODE POLICEPRECINCT CITYCOUNCILDISTRICT COMMUNITYDISTRICT
## 1 11208 75 37 305
## 2 11213 71 35 309
## 3 NA NA NA NA
## 4 10467 49 12 211
## 5 11210 70 45 314
## 6 11210 70 45 314
## COMMUNITYSCHOOLDISTRICT CONGRESSIONALDISTRICT ALARM_SOURCE_DESCRIPTION_TX
## 1 19 7 EMS
## 2 17 9 CLASS-3
## 3 NA NA EMS-911
## 4 11 15 EMS
## 5 22 9 EMS
## 6 22 9 EMS
## ALARM_LEVEL_INDEX_DESCRIPTION HIGHEST_ALARM_LEVEL
## 1 Initial Alarm First Alarm
## 2 Initial Alarm First Alarm
## 3 Initial Alarm First Alarm
## 4 DEFAULT RECORD First Alarm
## 5 DEFAULT RECORD First Alarm
## 6 DEFAULT RECORD First Alarm
## INCIDENT_CLASSIFICATION INCIDENT_CLASSIFICATION_GROUP
## 1 Medical - No PT Contact EMS is Onscene Medical Emergencies
## 2 Hospital Fire Structural Fires
## 3 Vehicle Accident - Other NonMedical Emergencies
## 4 Medical - EMS Link 10-91 Medical Emergencies
## 5 Medical - EMS Link 10-91 Medical Emergencies
## 6 Medical - EMS Link 10-91 Medical Emergencies
## DISPATCH_RESPONSE_SECONDS_QY FIRST_ASSIGNMENT_DATETIME
## 1 7 09/05/2023 02:19:12 PM
## 2 95 09/05/2023 02:21:11 PM
## 3 41 09/05/2023 02:20:25 PM
## 4 298 09/05/2023 02:25:59 PM
## 5 25 09/05/2023 02:21:52 PM
## 6 350 09/05/2023 02:28:25 PM
## FIRST_ACTIVATION_DATETIME FIRST_ON_SCENE_DATETIME INCIDENT_CLOSE_DATETIME
## 1 09/05/2023 02:19:26 PM 09/05/2023 02:25:23 PM 09/05/2023 03:03:15 PM
## 2 09/05/2023 02:21:33 PM 09/05/2023 02:23:21 PM 09/05/2023 02:34:18 PM
## 3 09/05/2023 02:20:35 PM 09/05/2023 02:26:22 PM 09/05/2023 04:13:32 PM
## 4 09/05/2023 02:26:04 PM 09/05/2023 02:34:23 PM
## 5 09/05/2023 02:22:08 PM 09/05/2023 02:28:07 PM
## 6 09/05/2023 02:29:09 PM
## VALID_DISPATCH_RSPNS_TIME_INDC VALID_INCIDENT_RSPNS_TIME_INDC
## 1 N Y
## 2 N Y
## 3 N Y
## 4 N N
## 5 N N
## 6 N N
## INCIDENT_RESPONSE_SECONDS_QY INCIDENT_TRAVEL_TM_SECONDS_QY
## 1 378 371
## 2 224 129
## 3 398 357
## 4 NA NA
## 5 NA NA
## 6 NA NA
## ENGINES_ASSIGNED_QUANTITY LADDERS_ASSIGNED_QUANTITY
## 1 1 0
## 2 3 2
## 3 2 3
## 4 1 0
## 5 1 0
## 6 1 0
## OTHER_UNITS_ASSIGNED_QUANTITY
## 1 0
## 2 1
## 3 1
## 4 0
## 5 0
## 6 0
Use dfSummary from summarytool in order to have a complete and clear sumamry of the dataset.
print(dfSummary(fire_data,
plain.ascii = FALSE,
style = "multiline",
headings = FALSE,
graph.magnif = 0.8,
valid.col = FALSE),
method = 'render')
Now we rename all the columns in order to be smaller whenever we plot graphs.
fire_data <- fire_data %>%
rename(id = STARFIRE_INCIDENT_ID, datetime = INCIDENT_DATETIME, al_borough = ALARM_BOX_BOROUGH,
al_number = ALARM_BOX_NUMBER,al_location = ALARM_BOX_LOCATION, inc_borough = INCIDENT_BOROUGH,
zipcode = ZIPCODE, pol_prec = POLICEPRECINCT, city_con_dist = CITYCOUNCILDISTRICT,
commu_dist = COMMUNITYDISTRICT, commu_sc_dist = COMMUNITYSCHOOLDISTRICT,
cong_dist = CONGRESSIONALDISTRICT, al_source_desc = ALARM_SOURCE_DESCRIPTION_TX,
al_index_desc = ALARM_LEVEL_INDEX_DESCRIPTION, highest_al_level = HIGHEST_ALARM_LEVEL,
inc_class = INCIDENT_CLASSIFICATION, inc_class_group = INCIDENT_CLASSIFICATION_GROUP,
first_ass_datetime = FIRST_ASSIGNMENT_DATETIME, first_act_datetime = FIRST_ACTIVATION_DATETIME,
first_onscene_datetime = FIRST_ON_SCENE_DATETIME, inc_close_datetime = INCIDENT_CLOSE_DATETIME,
disp_resp_sec_qy = DISPATCH_RESPONSE_SECONDS_QY, disp_resp_sec_indc = VALID_DISPATCH_RSPNS_TIME_INDC,
inc_resp_sec_qy = INCIDENT_RESPONSE_SECONDS_QY, inc_resp_sec_indc = VALID_INCIDENT_RSPNS_TIME_INDC,
inc_travel_sec_qy = INCIDENT_TRAVEL_TM_SECONDS_QY,
engines_assigned = ENGINES_ASSIGNED_QUANTITY,
ladders_assigned = LADDERS_ASSIGNED_QUANTITY, others_units_assigned = OTHER_UNITS_ASSIGNED_QUANTITY)
As we can see from the summary there are many NA values, and many predictors that are as characters and not factors. In this step we will convert the characters predictors as factors merging the values that appear less in the dataset, so we do no have many values that have low frequency in our dataset.
In addition we will add he predictor for the day_number, a factorial predictor to indicate in the incident day is a week day or not dat_type and a factorial predictor time_of_day that indicates the range of time whenever the incident happens, so Night (if the hour is between 0 and 6), Morning (if the hour is between 6 and 12), Afternoon (if the hour is between 12 and 18), Evening (if the hour is between 18 and 24).
Since we are dealing with datetime we also check if the differences (inc_resp_sec_qy, inc_travel_sec_qy and disp_resp_sec_qy) are actually corrects, if not we replace them with the correct one
# set factorial
fire_data$inc_borough <- as.factor(fire_data$inc_borough)
fire_data$al_borough <- as.factor(fire_data$al_borough)
fire_data$al_source_desc <- as.factor(fire_data$al_source_desc)
fire_data$al_index_desc <- as.factor(fire_data$al_index_desc)
fire_data$highest_al_level <- as.factor(fire_data$highest_al_level)
fire_data$disp_resp_sec_indc <- as.factor(fire_data$disp_resp_sec_indc)
levels(fire_data$disp_resp_sec_indc)<- c("N", "Y")
fire_data$inc_resp_sec_indc <- as.factor(fire_data$inc_resp_sec_indc)
levels(fire_data$inc_resp_sec_indc)<- c("N", "Y")
fire_data$inc_class_group <- as.factor(fire_data$inc_class_group)
fire_data$inc_class <- as.factor(fire_data$inc_class)
Moreover we note that the maximum level of the time indicator is very high to be considered as seconds so we decided to scale the two indicators in minutes.
summary(fire_data %>% select(inc_resp_sec_qy, inc_travel_sec_qy, disp_resp_sec_qy))
## inc_resp_sec_qy inc_travel_sec_qy disp_resp_sec_qy
## Min. : 18.0 Min. : 0.0 Min. : 2.00
## 1st Qu.: 265.0 1st Qu.: 233.0 1st Qu.: 7.00
## Median : 334.0 Median : 301.0 Median : 19.00
## Mean : 380.7 Mean : 340.5 Mean : 39.96
## 3rd Qu.: 426.0 3rd Qu.: 392.0 3rd Qu.: 40.00
## Max. :7130.0 Max. :7122.0 Max. :9023.00
## NA's :14112 NA's :14112
# scaling
fire_data$inc_resp_sec_qy <- fire_data$inc_resp_sec_qy / 60
fire_data$inc_travel_sec_qy <- fire_data$inc_travel_sec_qy / 60
fire_data$disp_resp_sec_qy <- fire_data$disp_resp_sec_qy / 60
# renaming both quantity and indicator predictors for the two datetime
fire_data <- fire_data %>% rename(inc_resp_min_qy = inc_resp_sec_qy, inc_travel_min_qy = inc_travel_sec_qy, disp_resp_min_qy = disp_resp_sec_qy, # quantity
inc_resp_min_indc = inc_resp_sec_indc, disp_resp_min_indc = disp_resp_sec_indc) # indicator
Here we create the time_of_day and is_weekend
# Process datetime column
fire_data$datetime <- mdy_hms(fire_data$datetime)
fire_data$first_ass_datetime <- mdy_hms(fire_data$first_ass_datetime)
fire_data$first_act_datetime <- mdy_hms(fire_data$first_act_datetime)
fire_data$first_onscene_datetime <- mdy_hms(fire_data$first_onscene_datetime)
fire_data$inc_close_datetime <- mdy_hms(fire_data$inc_close_datetime)
# checking if the differeces are well computed if not change with the correct one
if (!identical(fire_data$inc_resp_min_qy, as.numeric(difftime(fire_data$first_onscene_datetime, fire_data$datetime, units="mins")))){
fire_data$inc_resp_min_qy <- as.numeric(difftime(fire_data$first_onscene_datetime, fire_data$datetime, units="mins"))
}
if (!identical(fire_data$inc_travel_min_qy, as.numeric(difftime(fire_data$first_onscene_datetime, fire_data$first_ass_datetime, units="mins")))){
fire_data$inc_travel_min_qy <- as.numeric(difftime(fire_data$first_onscene_datetime, fire_data$first_ass_datetime, units="mins"))
}
if (!identical(fire_data$disp_resp_min_qy, as.numeric(difftime(fire_data$first_ass_datetime, fire_data$datetime, units="mins")))){
fire_data$disp_resp_min_qy <- as.numeric(difftime(fire_data$first_ass_datetime, fire_data$datetime, units="mins"))
}
# creating day_type
fire_data$day_type <- as.factor(ifelse(weekdays(fire_data$datetime) %in% c("sabato", "domenica"), "Weekend", "Weekday"))
# creating ticket_time
fire_data$ticket_time <- as.numeric(difftime(fire_data$inc_close_datetime, fire_data$datetime, units="mins"))
# creating time_of_day
fire_data$time_of_day <- cut(
hour(fire_data$datetime),
breaks = c(0, 6, 12, 18, 24),
labels = c("Night", "Morning", "Afternoon", "Evening"),
include.lowest = TRUE,
right = TRUE
)
fire_data$datetime <- NULL
table(fire_data$time_of_day)
##
## Night Morning Afternoon Evening
## 8521 13270 16499 11710
ggplot(data=fire_data %>%
group_by(time_of_day) %>%
summarise(incident_number = n()),
aes(x=time_of_day, y=incident_number)) +
geom_bar(stat="identity", position=position_dodge()) +
geom_text(aes(label=incident_number), vjust=1.6, color="white", position = position_dodge(0.9), size=3.5) +
labs(title = "Time of the Day - Incident Count", x = "Time of the Day", y = "Incident Count") +
theme_grey()
From this we can see that the higher number of fire incident is registered from 12 PM to 18 PM, whereas the lower number of fire incident happened from the 00 AM to 06 AM.
day_type_table <- table(fire_data$day_type)
day_type_table[1] <- day_type_table[1] / 5
day_type_table[2] <- day_type_table[2] / 2
day_type_table
##
## Weekday Weekend
## 7296.4 6759.0
And in proportion we can see that on average there is an higher number of fire incident on the week day respect to the week end days.
Now regarding the assigned untis we decided to add a summary predictor that include the sum of all the three assigned units predictors.
fire_data$total_assigned_unit <- fire_data$engines_assigned + fire_data$ladders_assigned + fire_data$others_units_assigned
Rename the factor levels for the inc_borough and predictors
fire_data <- fire_data %>% mutate(inc_borough = recode_factor(
inc_borough, "BRONX" = "Bronx", "BROOKLYN" = "Brooklyn", "MANHATTAN" = "Manhattan",
"QUEENS" = "Queens", "RICHMOND / STATEN ISLAND" = "Staten Island"),
al_borough = recode_factor(
al_borough, "BRONX" = "Bronx", "BROOKLYN" = "Brooklyn", "MANHATTAN" = "Manhattan",
"QUEENS" = "Queens", "RICHMOND / STATEN ISLAND" = "Staten Island"))
At this point we merge some possible value from factorial predictors to make the space of possible choice smaller.
Here we merge the following factorial values of highest_al_level: Second Alarm and Third Alarm into 2nd-3rd Alarm.
# highest_al_level
fire_data$highest_alarm_lev_new <- fire_data$highest_al_level
levels(fire_data$highest_alarm_lev_new) <- list(
"All Hands Working" = "All Hands Working",
"First Alarm" = "First Alarm",
"2nd-3rd Alarm" = c("Second Alarm", "Third Alarm")
)
print(ctable(fire_data$highest_al_level, fire_data$highest_alarm_lev_new, prop = 'n', totals = FALSE, headings = FALSE), method = 'render')
|
highest_alarm_lev_new |
|
highest_al_level
|
All Hands
Working |
First Alarm |
2nd-3rd
Alarm |
|
All Hands Working
|
100
|
0
|
0
|
|
First Alarm
|
0
|
49891
|
0
|
|
Second Alarm
|
0
|
0
|
8
|
|
Third Alarm
|
0
|
0
|
1
|
Generated by summarytools 1.0.1 (R version 4.2.1)
2024-01-10
fire_data$highest_al_level <- fire_data$highest_alarm_lev_new
fire_data$highest_alarm_lev_new <- NULL
Here we merge the following factorial values of al_index_desc: Second Alarm, Third Alarm, 7-5 (All Hands Alarm), 10-76 & 10-77 Signal (Notification Hi-Rise Fire) and 10-75 Signal (Request for all hands alarm) into Others.
# al_index_desc
fire_data$alarm_level_idx_new <- fire_data$al_index_desc
levels(fire_data$alarm_level_idx_new) <- list(
"DEFAULT RECORD" = "DEFAULT RECORD",
"Initial Alarm" = "Initial Alarm",
"Others" = c("Second Alarm", "Third Alarm", "7-5 (All Hands Alarm)",
"10-76 & 10-77 Signal (Notification Hi-Rise Fire)",
"10-75 Signal (Request for all hands alarm)")
)
print(ctable(fire_data$al_index_desc, fire_data$alarm_level_idx_new, prop = 'n', totals = FALSE, headings = FALSE), method = 'render')
|
alarm_level_idx_new |
|
al_index_desc
|
DEFAULT
RECORD |
Initial
Alarm |
Others |
|
10-75 Signal (Request for all hands alarm)
|
0
|
0
|
13
|
|
10-76 & 10-77 Signal (Notification Hi-Rise Fire)
|
0
|
0
|
3
|
|
7-5 (All Hands Alarm)
|
0
|
0
|
100
|
|
DEFAULT RECORD
|
17313
|
0
|
0
|
|
Initial Alarm
|
0
|
32562
|
0
|
|
Second Alarm
|
0
|
0
|
8
|
|
Third Alarm
|
0
|
0
|
1
|
Generated by summarytools 1.0.1 (R version 4.2.1)
2024-01-10
fire_data$al_index_desc <- fire_data$alarm_level_idx_new
fire_data$alarm_level_idx_new <- NULL
Here we merge the following factorial values of al_source_desc: 911, 911TEXT, VERBAL, BARS, ERS, ERS-NC and SOL into Others.
fire_data$alarm_source_desc_new <- fire_data$al_source_desc
levels(fire_data$alarm_source_desc_new) <- list(
"PHONE" = "PHONE",
"EMS" = "EMS",
"EMS-911" = "EMS-911",
"CLASS-3" = "CLASS-3",
"Others" = c("911", "911TEXT", "VERBAL", "BARS", "ERS", "ERS-NC", "SOL")
)
print(ctable(fire_data$al_source_desc, fire_data$alarm_source_desc_new, prop = 'n', totals = FALSE, headings = FALSE), method = 'render')
|
alarm_source_desc_new |
|
al_source_desc
|
PHONE |
EMS |
EMS-911 |
CLASS-3 |
Others |
|
911
|
0
|
0
|
0
|
0
|
302
|
|
911TEXT
|
0
|
0
|
0
|
0
|
14
|
|
BARS
|
0
|
0
|
0
|
0
|
1
|
|
CLASS-3
|
0
|
0
|
0
|
5025
|
0
|
|
EMS
|
0
|
17178
|
0
|
0
|
0
|
|
EMS-911
|
0
|
0
|
10520
|
0
|
0
|
|
ERS
|
0
|
0
|
0
|
0
|
777
|
|
ERS-NC
|
0
|
0
|
0
|
0
|
1
|
|
PHONE
|
15146
|
0
|
0
|
0
|
0
|
|
SOL
|
0
|
0
|
0
|
0
|
5
|
|
VERBAL
|
0
|
0
|
0
|
0
|
1031
|
Generated by summarytools 1.0.1 (R version 4.2.1)
2024-01-10
fire_data$al_source_desc <- fire_data$alarm_source_desc_new
fire_data$alarm_source_desc_new <- NULL
View again the dataset summary to see the applied changes.
print(dfSummary(fire_data,
plain.ascii = FALSE,
style = "multiline",
headings = FALSE,
graph.magnif = 0.8,
valid.col = FALSE),
method = 'render')
Dealing with invalid values
The next step is to deal invalid values and delete some un-useful predictors.
First of all we saw the possibility that al_borough and inc_borough represent the same column, let’s chek it.
identical(fire_data$al_borough, fire_data$inc_borough)
## [1] TRUE
The column `al_borough and inc_borough have the same sequence of values, so we can delete one of the two.
fire_data <- fire_data %>% select(-c(al_borough))
Then we say that all observation in the dataset have the disp_resp_min_indc equal to N, let’s check again and in affermative case then we can delete both columns.
summary(fire_data$disp_resp_min_indc)
## N Y
## 50000 0
All our observations have non valid disp_resp_min_indc so we can delete both the column indicator and the respective column quantity disp_resp_min_qy
fire_data <- fire_data %>% select(-c(disp_resp_min_indc, disp_resp_min_qy))
Now we do a quick check also on the other indicator variable inc_resp_min_indc
summary(fire_data$inc_resp_min_indc)
## N Y
## 17036 32964
But here we have some observations with valid inc_resp_min_indc, and we will consider only the valid one deleting the one that has a non valid attribute.
However before doing that let’s see the distribution of inc_resp_min_qy around the borough.
ggplot(data=fire_data %>% group_by(inc_borough, inc_resp_min_indc) %>% summarise(incident_number = n()),
aes(x=inc_borough, y=incident_number, fill=inc_resp_min_indc)) +
geom_bar(stat="identity", position=position_dodge()) +
geom_text(aes(label=incident_number), vjust=1.6, color="white",
position = position_dodge(0.9), size=3.5) +
scale_fill_brewer(palette="Paired") +
labs(title = "Incident Count - Borouh - Valid Response Time in Minutes", x = "Borough", y = "Incident Number", fill = "Valid Response\n Time in Minutes") +
theme_gray()
## `summarise()` has grouped output by 'inc_borough'. You can override using the
## `.groups` argument.
We can see that the number of fire incident is higher for the valid response time in minutes but, it is much interesting observe the rateo between the valid and the non valid.
And to the rateo of valid inc_resp_min_indc in each borough is:
rateo_inc_resp_min_indc <- fire_data %>%
group_by(inc_borough, inc_resp_min_indc) %>%
summarise(incident_number = n()) %>%
mutate(ratio=incident_number/sum(incident_number))
## `summarise()` has grouped output by 'inc_borough'. You can override using the
## `.groups` argument.
ggplot(rateo_inc_resp_min_indc, aes(fill=inc_resp_min_indc, y=ratio, x=inc_borough)) +
geom_bar(position="fill", stat="identity") +
geom_text(aes(label=scales::percent(ratio)), position=position_fill(vjust=0.5)) +
labs(title="Borough - Rateo Incident between Valid and Invalid",
x="Borough",
y="Rateo Incident between Valid and Invalid",
fill="Valid Response\nTime in Minutes")
And we can see that Staten Island has the higher number of incidents with valid inc_resp_min_indc , whereas Manhattan has the lower number, but remember that the former has the lowest number of fire incident and the latter has the higher number of incident.
Now we do an additional analysis to see if there is some find of relation between the inc_resp_min_indc and total_assigned_unit.
ggplot(fire_data, aes(total_assigned_unit, inc_resp_min_qy)) +
geom_point(aes(colour = inc_resp_min_indc))+
labs(title = "Total Assigned Units - Response Time In Minutes", x = "Total Assigned Units", y = "Response Time In Minutes", colour = "Valid Response\n Time in Minutes") +
theme_gray()
## Warning: Removed 14116 rows containing missing values (`geom_point()`).
We note that the majority of fire incident that had been assigned a single units has a high response time and the relative measure is not valid. Whereas for an higher number of total units the response time decrease and becomes valids.
ggplot(fire_data %>% filter(inc_resp_min_indc == "N")
, aes(total_assigned_unit, inc_resp_min_qy)) +
geom_point(aes(colour = inc_class_group)) +
labs(title = "Total Assigned Units - Response Time In Minutes - Incidnet Class Group", x = "Total Assigned Units", y = "Response Time In Minutes", colour = "Incident Class Groups") +
theme_gray()
## Warning: Removed 14112 rows containing missing values (`geom_point()`).
Regarding the incident class group around all the incidents with invalid response time had been assigned a single units as we discussed before, but in addition we found that are from the Medical Emergencies, whereas almost all the other incidents are from the NonMedical Emergencies.
# add an additional predictor
fire_data$tua_is_one <- as.factor(ifelse(fire_data$total_assigned_unit == 1, "Y", "N"))
tua_is_one <- fire_data %>%
filter(inc_resp_min_indc == "N", inc_class_group == "Medical Emergencies") %>%
group_by(inc_borough, tua_is_one) %>%
summarise(incident_number = n())
## `summarise()` has grouped output by 'inc_borough'. You can override using the
## `.groups` argument.
ggplot(data=tua_is_one,
aes(x=inc_borough, y=incident_number, fill=tua_is_one)) +
geom_bar(stat="identity", position=position_dodge()) +
geom_text(aes(label=incident_number), vjust=1.5, color="black",
position = position_dodge(0.9), size=3.5) +
scale_fill_brewer(palette="Set1") +
labs(title = "Total Assigned Units One or Not", x = "Borough", y = "Incident Count", fill = "Total Assigned\nUnits are One") +
theme_gray()
We have also added an additional factorial predictor tua_is_one to indicates if the total assigned units is equal to one or not.
Continuing we decide to analyse the type of Incident Class of the invalid incidents response time that had been assigned a single total units.
ggplot(data=fire_data %>%
filter(inc_resp_min_indc == "N", inc_class_group == "Medical Emergencies", tua_is_one == "Y") %>%
group_by(inc_class, inc_borough) %>%
summarise(incident_number = n()),
aes(x=inc_borough, y=incident_number, fill=inc_class)) +
geom_bar(stat="identity", position=position_dodge()) +
geom_text(aes(label=incident_number), vjust=1.6, color="black",
position = position_dodge(0.9), size=3) +
scale_fill_brewer(palette="Set1") +
labs(title = "Borough - Incident Counts - Incident Class -- for Total Assigned Units equal to 1", x = "Borough", y = "Incident Counts", fill = "Incident Class Group") +
theme_grey()
## `summarise()` has grouped output by 'inc_class'. You can override using the
## `.groups` argument.
And we found that the majority of the incident that respect these circumstances are mostly identified as Medical - EMS Link 10-91 and Medical - PD Link 10-91.
Thanks to the 10code site we found a description of the two emergency codes:
10-91 Medical Emergency EMS - Fire Unit Not Required - To be transmitted through borough dispatcher by the responding unit when the fire Unit is canceled enroute due to EMS on scene, or EMS downgrades the job to a segment that does not require a Fire Unit response. Note: This signal shall be used only for medical emergency incidents. EMS we are sure that stands for Emergency Medical Services.
10-91 Medical Emergency PD - Fire Unit Not Required - To be transmitted through borough dispatcher by the responding unit when the fire Unit is canceled enroute due to PD on scene, or PD downgrades the job to a segment that does not require a Fire Unit response. Note: This signal shall be used only for medical emergency incidents. PD we think that stands for Police Department.
Now we can look for the NonMedical Emergencies by first see the distribution of its incident class.
print(fire_data %>%
filter(inc_resp_min_indc == "N", inc_class_group == "NonMedical Emergencies") %>%
group_by(inc_class) %>%
summarise(incident_number = n()))
## # A tibble: 24 × 2
## inc_class incident_number
## <fct> <int>
## 1 Alarm System - Defective 10
## 2 Alarm System - Testing 22
## 3 Alarm System - Unnecessary 110
## 4 Assist Civilian - Non-Medical 828
## 5 Carbon Monoxide - Code 1 - Investigation 25
## 6 Carbon Monoxide - Code 2 - Incident (1-9 ppm) 4
## 7 Carbon Monoxide - Code 3 - Emergency (over 9 ppm) 4
## 8 Defective Oil Burner 5
## 9 Downed Tree 28
## 10 Elevator Emergency - Occupied 104
## # ℹ 14 more rows
ggplot(data=fire_data %>%
filter(inc_resp_min_indc == "N", inc_class_group == "NonMedical Emergencies", inc_class == "Assist Civilian - Non-Medical") %>%
group_by(inc_borough) %>%
summarise(incident_number = n()),
aes(x=inc_borough, y=incident_number)) +
geom_bar(stat="identity", position=position_dodge()) +
geom_text(aes(label=incident_number), vjust=1.6, color="white", position = position_dodge(0.9), size=3.5) +
#scale_fill_brewer(palette="Paired") +
labs(title = "Incident Count - Borouh - Valid Response Time in Second", x = "Borough", y = "Incident Count") +
theme_grey()
And we found that the majority of non valid inc_resp_min_indc that are Non-Medical Emergency are from the incident class equal to Assist Civilian - Non-Medical.
For stake of consistency we will consider only the valid observations that have inc_resp_min_indc == "Y".
fire_data <- fire_data %>% filter(inc_resp_min_indc == "Y")
dim(fire_data)
## [1] 32964 30
Now we want to know how many inc_class are summarized in each inc_class_group, to be sure that each inc_class_group is referred to a single inc_class.
print(ctable(fire_data$inc_class, fire_data$inc_class_group, totals = FALSE, headings = FALSE), method = 'render')
|
inc_class_group |
|
inc_class
|
Medical
Emergencies |
Medical MFAs |
NonMedical
Emergencies |
NonMedical
MFAs |
NonStructura
l Fires |
Structural
Fires |
|
Abandoned Derelict Vehicle Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
6 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Alarm System - Defective
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
377 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Alarm System - Testing
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
706 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Alarm System - Unnecessary
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
2735 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Assist Civilian - Non-Medical
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
3312 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Automobile Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
101 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Brush Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
24 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Carbon Monoxide - Code 1 - Investigation
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
788 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Carbon Monoxide - Code 2 - Incident (1-9 ppm)
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
129 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Carbon Monoxide - Code 3 - Emergency (over 9 ppm)
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
88 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Carbon Monoxide - Code 4 - No Detector Activation
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
8 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Church Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
10 |
( |
100.0% |
) |
|
Defective Oil Burner
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
34 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Demolition Debris or Rubbish Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
272 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Downed Tree
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
280 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Elevator Emergency - Occupied
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1850 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Elevator Emergency - Unoccupied
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
708 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Factory Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1 |
( |
100.0% |
) |
|
Hospital Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
18 |
( |
100.0% |
) |
|
Manhole Fire - Blown Cover
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
9 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Manhole Fire - Other
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
55 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Manhole Fire - Seeping Smoke
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
104 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Maritime Emergency
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Maritime Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - Assist Civilian
|
27 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - Breathing / Ill or Sick
|
4779 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - EMS Link 10-91
|
1096 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - No PT Contact EMS is Onscene
|
4285 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - PD Link 10-91
|
868 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - Serious Life Threatening
|
366 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical - Victim Deceased
|
287 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical MFA - EMS Link
|
0 |
( |
0.0% |
) |
87 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Medical MFA - PD Link
|
0 |
( |
0.0% |
) |
77 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Multiple Dwelling 'A' - Compactor fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
4 |
( |
100.0% |
) |
|
Multiple Dwelling 'A' - Food on the stove fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
519 |
( |
100.0% |
) |
|
Multiple Dwelling 'A' - Other fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
168 |
( |
100.0% |
) |
|
Multiple Dwelling 'B' Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
85 |
( |
100.0% |
) |
|
Non-Medical 10-91 (Unnecessary Alarm)
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
102 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Non-Medical MFA - ERS
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
586 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Non-Medical MFA - ERS No Contact
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Non-Medical MFA - Phone
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
701 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Non-Medical MFA - Private Fire Alarm
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
223 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Non-Medical MFA - Verbal
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
7 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Odor - Other Smoke
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
166 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Odor - Other Than Smoke
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1317 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Other Commercial Building Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
184 |
( |
100.0% |
) |
|
Other Public Building Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
4 |
( |
100.0% |
) |
|
Other Transportation Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
14 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Private Dwelling Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
412 |
( |
100.0% |
) |
|
Remove Civilian - Non-Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
27 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
School Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
31 |
( |
100.0% |
) |
|
Sprinkler System - Activated
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
6 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Sprinkler System - Malfunction
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
41 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Sprinkler System - Working on System
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
28 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Store Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
9 |
( |
100.0% |
) |
|
Transit System - NonStructural
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
59 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
|
Transit System - Structural
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1 |
( |
100.0% |
) |
|
Transit System Emergency
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
18 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Undefined Emergency
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
71 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Under Contruction / Vacant Fire
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1 |
( |
100.0% |
) |
|
Utility Emergency - Electric
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
595 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Utility Emergency - Gas
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1335 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Utility Emergency - Steam
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
137 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Utility Emergency - Undefined
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
4 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Utility Emergency - Water
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1157 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Vehicle Accident - Other
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
1443 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
|
Vehicle Accident - With Extrication
|
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
21 |
( |
100.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
0 |
( |
0.0% |
) |
Generated by summarytools 1.0.1 (R version 4.2.1)
2024-01-10
As we can see from the upper table all the inc_class_group have a unique set of values.
At this point to be more clear we display each main class with each respective sub-class.
for (variable in levels(fire_data$inc_class_group)) {
non_zero_table <- table(subset(fire_data, inc_class_group == variable)$inc_class)
cat(variable, "\n")
print(non_zero_table[non_zero_table != 0])
cat("\n")
}
## Medical Emergencies
##
## Medical - Assist Civilian Medical - Breathing / Ill or Sick
## 27 4779
## Medical - EMS Link 10-91 Medical - No PT Contact EMS is Onscene
## 1096 4285
## Medical - PD Link 10-91 Medical - Serious Life Threatening
## 868 366
## Medical - Victim Deceased
## 287
##
## Medical MFAs
##
## Medical MFA - EMS Link Medical MFA - PD Link
## 87 77
##
## NonMedical Emergencies
##
## Alarm System - Defective
## 377
## Alarm System - Testing
## 706
## Alarm System - Unnecessary
## 2735
## Assist Civilian - Non-Medical
## 3312
## Carbon Monoxide - Code 1 - Investigation
## 788
## Carbon Monoxide - Code 2 - Incident (1-9 ppm)
## 129
## Carbon Monoxide - Code 3 - Emergency (over 9 ppm)
## 88
## Carbon Monoxide - Code 4 - No Detector Activation
## 8
## Defective Oil Burner
## 34
## Downed Tree
## 280
## Elevator Emergency - Occupied
## 1850
## Elevator Emergency - Unoccupied
## 708
## Non-Medical 10-91 (Unnecessary Alarm)
## 102
## Odor - Other Smoke
## 166
## Odor - Other Than Smoke
## 1317
## Remove Civilian - Non-Fire
## 27
## Sprinkler System - Activated
## 6
## Sprinkler System - Malfunction
## 41
## Sprinkler System - Working on System
## 28
## Transit System Emergency
## 18
## Undefined Emergency
## 71
## Utility Emergency - Electric
## 595
## Utility Emergency - Gas
## 1335
## Utility Emergency - Steam
## 137
## Utility Emergency - Undefined
## 4
## Utility Emergency - Water
## 1157
## Vehicle Accident - Other
## 1443
## Vehicle Accident - With Extrication
## 21
##
## NonMedical MFAs
##
## Non-Medical MFA - ERS Non-Medical MFA - ERS No Contact
## 586 1
## Non-Medical MFA - Phone Non-Medical MFA - Private Fire Alarm
## 701 223
## Non-Medical MFA - Verbal
## 7
##
## NonStructural Fires
##
## Abandoned Derelict Vehicle Fire Automobile Fire
## 6 101
## Brush Fire Demolition Debris or Rubbish Fire
## 24 272
## Manhole Fire - Blown Cover Manhole Fire - Other
## 9 55
## Manhole Fire - Seeping Smoke Other Transportation Fire
## 104 14
## Transit System - NonStructural
## 59
##
## Structural Fires
##
## Church Fire
## 10
## Factory Fire
## 1
## Hospital Fire
## 18
## Multiple Dwelling 'A' - Compactor fire
## 4
## Multiple Dwelling 'A' - Food on the stove fire
## 519
## Multiple Dwelling 'A' - Other fire
## 168
## Multiple Dwelling 'B' Fire
## 85
## Other Commercial Building Fire
## 184
## Other Public Building Fire
## 4
## Private Dwelling Fire
## 412
## School Fire
## 31
## Store Fire
## 9
## Transit System - Structural
## 1
## Under Contruction / Vacant Fire
## 1
NA Patterns?
At this point is essential to deal with NA values, trying to find the presence of possible relation with predictors. First things first let’s recap the number of NA values for each columns that we have at the moment.
colSums(is.na(fire_data))
## id al_number al_location
## 0 0 0
## inc_borough zipcode pol_prec
## 0 2197 2197
## city_con_dist commu_dist commu_sc_dist
## 2197 2197 2198
## cong_dist al_source_desc al_index_desc
## 2197 0 0
## highest_al_level inc_class inc_class_group
## 0 0 0
## first_ass_datetime first_act_datetime first_onscene_datetime
## 0 41 0
## inc_close_datetime inc_resp_min_indc inc_resp_min_qy
## 0 0 0
## inc_travel_min_qy engines_assigned ladders_assigned
## 0 4 4
## others_units_assigned day_type ticket_time
## 4 0 0
## time_of_day total_assigned_unit tua_is_one
## 0 4 4
Checking the location predictors
Here we will check if there is a pattern on the absence of values in the following predictors: zipcode, pol_prec, city_con_dist, commu_dist, commu_sc_dist and cong_dist.
na_locations <- fire_data %>%
filter(is.na(zipcode) | is.na(pol_prec) | is.na(city_con_dist) | is.na(commu_dist) | is.na(commu_sc_dist) | is.na(cong_dist))
ggplot(data=na_locations %>%
group_by(inc_class_group, inc_borough) %>%
summarise(incident_number = n()),
aes(x=inc_borough, y=incident_number, fill=inc_class_group)) + geom_bar(stat="identity", position=position_dodge()) +
geom_text(aes(label=incident_number), vjust=1.6, color="black",
position = position_dodge(0.9), size=3.5) +
#scale_fill_brewer(palette="Paired") +
labs(title = "NA location", x = "Borough", y = "Incident Count", fill = "Incident Class Group") +
theme_grey()
## `summarise()` has grouped output by 'inc_class_group'. You can override using
## the `.groups` argument.
By the Bar Chart we note that the majority of observations that have at least one of the location predictors to NA are of the incident class group NonMedical Emergency, Non Medical MFAs and Medical Emergencies
table(na_locations$inc_borough) / table(fire_data$inc_borough)
##
## Bronx Brooklyn Manhattan Queens Staten Island
## 0.07104538 0.04694547 0.07662157 0.07700328 0.07523697
table(na_locations$inc_class_group) / table(fire_data$inc_class_group)
##
## Medical Emergencies Medical MFAs NonMedical Emergencies
## 0.03988726 0.10975610 0.05691243
## NonMedical MFAs NonStructural Fires Structural Fires
## 0.40447958 0.14906832 0.00552868
Moreover around the 40% of the whole incident that are of the incident class group NonMedical MFAs have at least one of the location columns to NA. Let’s investigate.
fd_nm_mfa_cl <- table(subset(fire_data, inc_class_group == "NonMedical MFAs")$inc_class)
fd_nm_mfa_bro <- table(subset(fire_data, inc_class_group == "NonMedical MFAs")$inc_borough)
fd_nm_mfa_cl <- fd_nm_mfa_cl[fd_nm_mfa_cl != 0]
fd_nm_mfa_cl
##
## Non-Medical MFA - ERS Non-Medical MFA - ERS No Contact
## 586 1
## Non-Medical MFA - Phone Non-Medical MFA - Private Fire Alarm
## 701 223
## Non-Medical MFA - Verbal
## 7
In the original dataset this is the distribution of inc_class for the NonMedical MF
na_nm_mfa_cl <- table(subset(na_locations, inc_class_group == "NonMedical MFAs")$inc_class)
na_nm_mfa_bro <- table(subset(na_locations, inc_class_group == "NonMedical MFAs")$inc_borough)
na_nm_mfa_cl <- na_nm_mfa_cl[names(fd_nm_mfa_cl)]
na_nm_mfa_cl
##
## Non-Medical MFA - ERS Non-Medical MFA - ERS No Contact
## 573 1
## Non-Medical MFA - Phone Non-Medical MFA - Private Fire Alarm
## 38 2
## Non-Medical MFA - Verbal
## 0
na_nm_mfa_cl / fd_nm_mfa_cl
##
## Non-Medical MFA - ERS Non-Medical MFA - ERS No Contact
## 0.97781570 1.00000000
## Non-Medical MFA - Phone Non-Medical MFA - Private Fire Alarm
## 0.05420827 0.00896861
## Non-Medical MFA - Verbal
## 0.00000000
So the 97% of all the Non-Medical MFA - ERS observations in the entire dataset have one of the location attribute equal to NA
na_nm_mfa_bro / fd_nm_mfa_bro
##
## Bronx Brooklyn Manhattan Queens Staten Island
## 0.4676056 0.3075221 0.3894472 0.3733333 0.7954545
And from here we can see that about the 78% of the observations that are NonMedical - MFAs that have at least one district column attribute to NA are from the RICHMOND / STATEN ISLAND. Also BRONX has about half of the NonMedical - MFAs observations that have at least one district column to NA.
Checking the assigned units predictors
print(fire_data %>%
filter(is.na(engines_assigned) | is.na(ladders_assigned) | is.na(others_units_assigned)) %>%
group_by(inc_borough, inc_class)) %>%
summarise(incident_count = n())
## # A tibble: 4 × 30
## # Groups: inc_borough, inc_class [4]
## id al_number al_location inc_borough zipcode pol_prec city_con_dist
## <chr> <int> <chr> <fct> <int> <int> <int>
## 1 230905-Q4545… 4545 53 AVE & 6… Queens 11378 104 30
## 2 230914-Q1014… 1014 CENTRAL AV… Queens 11691 101 31
## 3 230918-Q9643… 9643 JAMAICA AV… Queens 11418 102 29
## 4 230919-M0684… 684 1 AVE & E … Manhattan 10016 13 4
## # ℹ 23 more variables: commu_dist <int>, commu_sc_dist <int>, cong_dist <int>,
## # al_source_desc <fct>, al_index_desc <fct>, highest_al_level <fct>,
## # inc_class <fct>, inc_class_group <fct>, first_ass_datetime <dttm>,
## # first_act_datetime <dttm>, first_onscene_datetime <dttm>,
## # inc_close_datetime <dttm>, inc_resp_min_indc <fct>, inc_resp_min_qy <dbl>,
## # inc_travel_min_qy <dbl>, engines_assigned <int>, ladders_assigned <int>,
## # others_units_assigned <int>, day_type <fct>, ticket_time <dbl>, …
## `summarise()` has grouped output by 'inc_borough'. You can override using the
## `.groups` argument.
## # A tibble: 4 × 3
## # Groups: inc_borough [2]
## inc_borough inc_class incident_count
## <fct> <fct> <int>
## 1 Manhattan Vehicle Accident - Other 1
## 2 Queens Assist Civilian - Non-Medical 1
## 3 Queens Medical - No PT Contact EMS is Onscene 1
## 4 Queens Medical - PD Link 10-91 1
We can easily remove this observations.
Checking the first_act_datetime predictors
na_first_act_datetime <- fire_data %>% filter(is.na(first_act_datetime))
print(na_first_act_datetime %>% group_by(inc_class, inc_borough) %>% summarise(incident_count = n()))
## `summarise()` has grouped output by 'inc_class'. You can override using the
## `.groups` argument.
## # A tibble: 27 × 3
## # Groups: inc_class [15]
## inc_class inc_borough incident_count
## <fct> <fct> <int>
## 1 Alarm System - Unnecessary Brooklyn 2
## 2 Assist Civilian - Non-Medical Bronx 1
## 3 Assist Civilian - Non-Medical Brooklyn 4
## 4 Assist Civilian - Non-Medical Queens 1
## 5 Demolition Debris or Rubbish Fire Brooklyn 1
## 6 Downed Tree Manhattan 1
## 7 Downed Tree Queens 1
## 8 Downed Tree Staten Island 1
## 9 Elevator Emergency - Occupied Brooklyn 1
## 10 Elevator Emergency - Occupied Manhattan 1
## # ℹ 17 more rows
ggplot(data=na_first_act_datetime %>%
group_by(inc_class_group, inc_borough) %>%
summarise(incident_number = n()),
aes(x=inc_borough, y=incident_number, fill=inc_class_group)) + geom_bar(stat="identity", position=position_dodge()) +
labs(title = "NA First Act Date", x = "Borough", y = "Incident Count", fill = "Incident Class Group") +
theme_minimal()
## `summarise()` has grouped output by 'inc_class_group'. You can override using
## the `.groups` argument.
Seems to be random and thus there is no pattern that motivate the presence of NA values in first_act_datetime.
At this point we can omit the NA values.
fire_data_new <- na.omit(fire_data)
And the un-usefull predictors
fire_data_new <- fire_data_new %>% select(-c(zipcode, pol_prec, city_con_dist, commu_dist, al_location,
commu_sc_dist, cong_dist, first_ass_datetime, first_act_datetime,
first_onscene_datetime, inc_close_datetime, inc_resp_min_indc,
inc_class, id, al_number))
print(dfSummary(fire_data,
plain.ascii = FALSE,
style = "multiline",
headings = FALSE,
graph.magnif = 0.8,
valid.col = FALSE),
method = 'render')
Additional Data Visaulization
In this section we will have a look on additional data visualisation in order to better understand how the predictors behaves.
summary(fire_data_new)
## inc_borough al_source_desc al_index_desc
## Bronx :6406 PHONE :13465 DEFAULT RECORD: 2979
## Brooklyn :9280 EMS : 7369 Initial Alarm :27631
## Manhattan :7294 EMS-911: 4349 Others : 118
## Queens :6188 CLASS-3: 4817
## Staten Island:1560 Others : 728
##
## highest_al_level inc_class_group inc_resp_min_qy
## All Hands Working: 94 Medical Emergencies :11222 Min. : 0.350
## First Alarm :30626 Medical MFAs : 146 1st Qu.: 4.383
## 2nd-3rd Alarm : 8 NonMedical Emergencies:16471 Median : 5.467
## NonMedical MFAs : 904 Mean : 5.949
## NonStructural Fires : 547 3rd Qu.: 6.850
## Structural Fires : 1438 Max. :58.917
## inc_travel_min_qy engines_assigned ladders_assigned others_units_assigned
## Min. : 0.000 Min. : 0.000 Min. : 0.0000 Min. : 0.0000
## 1st Qu.: 3.883 1st Qu.: 1.000 1st Qu.: 0.0000 1st Qu.: 0.0000
## Median : 4.983 Median : 1.000 Median : 1.0000 Median : 0.0000
## Mean : 5.448 Mean : 1.146 Mean : 0.7795 Mean : 0.3899
## 3rd Qu.: 6.367 3rd Qu.: 1.000 3rd Qu.: 1.0000 3rd Qu.: 1.0000
## Max. :58.617 Max. :19.000 Max. :15.0000 Max. :32.0000
## day_type ticket_time time_of_day total_assigned_unit
## Weekday:22379 Min. : 1.083 Night : 4487 Min. : 1.000
## Weekend: 8349 1st Qu.: 12.850 Morning : 8370 1st Qu.: 1.000
## Median : 17.900 Afternoon:10627 Median : 1.000
## Mean : 23.027 Evening : 7244 Mean : 2.316
## 3rd Qu.: 25.954 3rd Qu.: 3.000
## Max. :601.717 Max. :66.000
## tua_is_one
## N:12940
## Y:17788
##
##
##
##
ggplot(fire_data_new,
aes(x = inc_borough, y = inc_resp_min_qy, color = inc_borough)) + # ggplot function
geom_boxplot()# +
#labs(title = "Borough - Incident Number - Alarm Source", x = "Borough", y = "Incident Number", color = "Alarm Source")
ggplot(fire_data_new,
aes(x = al_source_desc, y = inc_resp_min_qy, color = al_source_desc)) + # ggplot function
geom_boxplot()# +
#labs(title = "Borough - Incident Number - Alarm Source", x = "Borough", y = "Incident Number", color = "Alarm Source")
ggplot(fire_data_new,
aes(x = day_type, y = inc_resp_min_qy, color = day_type)) + # ggplot function
geom_boxplot()# +
#labs(title = "Borough - Incident Number - Alarm Source", x = "Borough", y = "Incident Number", color = "Alarm Source")
ggplot(fire_data_new,
aes(x = time_of_day, y = inc_resp_min_qy, color = time_of_day)) + # ggplot function
geom_boxplot()# +
#labs(title = "Borough - Incident Number - Alarm Source", x = "Borough", y = "Incident Number", color = "Alarm Source")
Maps Visualization
In this section we plot additional data visualization focus on the geographical visualization of the New York borough with relative predictors. In order to do so we load two additional datasets: 1. Alarm_Box_Locations.csv is a dataset that includes geographical informations about the alarm box, including latitude and longitude useful to plot points in a map. 2. fdny-firehouse-listing.csv is a dataset that includes the geographical informations of every firefighter stations in the NYC, including again latitude and longitude.
alarm_box_loc <- read.csv("datasets/Alarm_Box_Locations.csv")
# IF I WANT TO USE THIS I HAVE DO DELETE THE REMOVING STEP OF al_location UPPER
head(alarm_box_loc)
## BOROBOX BOX_TYPE LOCATION ZIP BOROUGH
## 1 B2653 ERS 3 AVE & 65 ST 11220 Brooklyn
## 2 Q7917 BARS WOODSIDE AVE & 69 ST 11377 Queens
## 3 B0801 ERS MYRTLE AVE & PALMETTO ST 11237 Brooklyn
## 4 B1046 ERS NEW YORK AVE & LEFFERTS AVE 11225 Brooklyn
## 5 B0109 ERS RIVER & NORTH 3 STS 11211 Brooklyn
## 6 R2465 BARS VINCENT AVE & COVERLY ST 10306 Staten Island
## COMMUNITYDISTICT CITYCOUNCIL LATITUDE LONGITUDE
## 1 BK07 38 40.63932 -74.02355
## 2 QN02 26 40.74269 -73.89565
## 3 QN05 34 40.69953 -73.91103
## 4 BK09 40 40.66253 -73.94791
## 5 BK01 33 40.71838 -73.96462
## 6 SI03 50 40.57084 -74.12511
## Location.Point
## 1 POINT (-74.02354939 40.63932033)
## 2 POINT (-73.89565167 40.7426855)
## 3 POINT (-73.9110349 40.69953211)
## 4 POINT (-73.94791393 40.66253364)
## 5 POINT (-73.96462115 40.71837562)
## 6 POINT (-74.12510919 40.57084247)
summary(alarm_box_loc)
## BOROBOX BOX_TYPE LOCATION ZIP
## Length:13008 Length:13008 Length:13008 Min. : 83
## Class :character Class :character Class :character 1st Qu.:10314
## Mode :character Mode :character Mode :character Median :11211
## Mean :10864
## 3rd Qu.:11369
## Max. :11697
## NA's :27
## BOROUGH COMMUNITYDISTICT CITYCOUNCIL LATITUDE
## Length:13008 Length:13008 Min. : 1.00 Min. :40.50
## Class :character Class :character 1st Qu.:19.00 1st Qu.:40.64
## Mode :character Mode :character Median :28.00 Median :40.71
## Mean :28.89 Mean :40.71
## 3rd Qu.:41.00 3rd Qu.:40.76
## Max. :51.00 Max. :40.91
## NA's :4
## LONGITUDE Location.Point
## Min. :-74.25 Length:13008
## 1st Qu.:-73.98 Class :character
## Median :-73.92 Mode :character
## Mean :-73.92
## 3rd Qu.:-73.84
## Max. :-73.70
##
firefighter_stations <- read.csv("datasets/fdny-firehouse-listing.csv")
head(firefighter_stations)
## FacilityName FacilityAddress
## 1 Engine 4/Ladder 15 42 South Street
## 2 Engine 10/Ladder 10 124 Liberty Street
## 3 Engine 6 49 Beekman Street
## 4 Engine 7/Ladder 1/Battalion 1/Manhattan Borough Command 100-104 Duane Street
## 5 Ladder 8 14 North Moore Street
## 6 Engine 9/Ladder 6 75 Canal Street
## Borough Postcode Latitude Longitude Community.Board Community.Council
## 1 Manhattan 10005 40.70347 -74.00754 1 1
## 2 Manhattan 10006 40.71007 -74.01252 1 1
## 3 Manhattan 10038 40.71005 -74.00525 1 1
## 4 Manhattan 10007 40.71546 -74.00594 1 1
## 5 Manhattan 10013 40.71976 -74.00668 1 1
## 6 Manhattan 10002 40.71521 -73.99290 3 1
## Census.Tract BIN BBL
## 1 7 1000867 1000350001
## 2 13 1075700 1000520022
## 3 1501 1001287 1000930030
## 4 33 1001647 1001500025
## 5 33 1002150 1001890035
## 6 16 1003898 1003000030
## NTA
## 1 Battery Park City-Lower Manhattan
## 2 Battery Park City-Lower Manhattan
## 3 Battery Park City-Lower Manhattan
## 4 SoHo-TriBeCa-Civic Center-Little Italy
## 5 SoHo-TriBeCa-Civic Center-Little Italy
## 6 Chinatown
summary(firefighter_stations)
## FacilityName FacilityAddress Borough Postcode
## Length:218 Length:218 Length:218 Min. :10001
## Class :character Class :character Class :character 1st Qu.:10304
## Mode :character Mode :character Mode :character Median :11103
## Mean :10784
## 3rd Qu.:11231
## Max. :11695
## NA's :5
## Latitude Longitude Community.Board Community.Council
## Min. :40.51 Min. :-74.24 Min. : 1.000 Min. : 1.00
## 1st Qu.:40.66 1st Qu.:-73.99 1st Qu.: 3.000 1st Qu.:12.00
## Median :40.72 Median :-73.94 Median : 6.000 Median :27.00
## Mean :40.72 Mean :-73.94 Mean : 7.075 Mean :25.63
## 3rd Qu.:40.77 3rd Qu.:-73.89 3rd Qu.:11.000 3rd Qu.:38.00
## Max. :40.89 Max. :-73.72 Max. :84.000 Max. :51.00
## NA's :5 NA's :5 NA's :5 NA's :5
## Census.Tract BIN BBL NTA
## Min. : 1 Min. :1000867 Min. :1.000e+09 Length:218
## 1st Qu.: 129 1st Qu.:2003268 1st Qu.:2.025e+09 Class :character
## Median : 275 Median :3064786 Median :3.025e+09 Mode :character
## Mean : 5950 Mean :2900421 Mean :2.850e+09
## 3rd Qu.: 800 3rd Qu.:4090228 3rd Qu.:4.033e+09
## Max. :157902 Max. :5154879 Max. :5.080e+09
## NA's :5 NA's :5 NA's :5
We now start with the firefighter stations dataset. By first making a copy of the fire_data_new and setting the borough from the firefighter_stations dataset to factor in order to be easily merged with the copied fire_data dataset.
# make a copy of the fire_data
fire_data_for_ffs <- fire_data_new
fire_data_for_ffs <- fire_data_for_ffs %>% rename(borough = inc_borough)
firefighter_stations$Borough <- as.factor(firefighter_stations$Borough)
firefighter_stations <- firefighter_stations %>% rename(borough = Borough)
# remove the na values from firefighter_stations
firefighter_stations <- na.omit(firefighter_stations)
Now we want to get the number of firefighter station for each borough.
stations_borough <- firefighter_stations %>%
group_by(borough) %>%
summarise(number_of_stations = n())
Now we want to get a summary of the incident count, the number of station and the incident per station of each borough in order to have a general view of the New York City situation.
count_inc_brough <- fire_data_for_ffs %>% group_by(borough) %>% summarise(incident_count = n())
stations_borough$incident_per_station <- round(count_inc_brough$incident_count / stations_borough$number_of_stations, digits = 3)
count_inc_brough <- merge(count_inc_brough, stations_borough, by="borough")
count_inc_brough
## borough incident_count number_of_stations incident_per_station
## 1 Bronx 6406 34 188.412
## 2 Brooklyn 9280 64 145.000
## 3 Manhattan 7294 47 155.191
## 4 Queens 6188 48 128.917
## 5 Staten Island 1560 20 78.000
Now we convert the firefighter_station data frame into a Spartial Data Frame to contains the geometry points.
firefighter_stations_sdf <- st_as_sf(firefighter_stations, coords = c("Longitude", "Latitude"), crs = 4326)
head(firefighter_stations_sdf)
## Simple feature collection with 6 features and 10 fields
## Geometry type: POINT
## Dimension: XY
## Bounding box: xmin: -74.01252 ymin: 40.70347 xmax: -73.9929 ymax: 40.71976
## Geodetic CRS: WGS 84
## FacilityName FacilityAddress
## 1 Engine 4/Ladder 15 42 South Street
## 2 Engine 10/Ladder 10 124 Liberty Street
## 3 Engine 6 49 Beekman Street
## 4 Engine 7/Ladder 1/Battalion 1/Manhattan Borough Command 100-104 Duane Street
## 5 Ladder 8 14 North Moore Street
## 6 Engine 9/Ladder 6 75 Canal Street
## borough Postcode Community.Board Community.Council Census.Tract BIN
## 1 Manhattan 10005 1 1 7 1000867
## 2 Manhattan 10006 1 1 13 1075700
## 3 Manhattan 10038 1 1 1501 1001287
## 4 Manhattan 10007 1 1 33 1001647
## 5 Manhattan 10013 1 1 33 1002150
## 6 Manhattan 10002 3 1 16 1003898
## BBL
## 1 1000350001
## 2 1000520022
## 3 1000930030
## 4 1001500025
## 5 1001890035
## 6 1003000030
## NTA
## 1 Battery Park City-Lower Manhattan
## 2 Battery Park City-Lower Manhattan
## 3 Battery Park City-Lower Manhattan
## 4 SoHo-TriBeCa-Civic Center-Little Italy
## 5 SoHo-TriBeCa-Civic Center-Little Italy
## 6 Chinatown
## geometry
## 1 POINT (-74.00754 40.70347)
## 2 POINT (-74.01252 40.71007)
## 3 POINT (-74.00525 40.71005)
## 4 POINT (-74.00594 40.71546)
## 5 POINT (-74.00668 40.71976)
## 6 POINT (-73.9929 40.71521)
Downloand of the geojson file
At this point we download the .geojson file that contaions all the geometry of each borough in order to have a cool maps visualization of NYC.
geojson_newyork <- geojson_read("datasets/NYC_BoroughBoundaries.geojson", what = "sp")
geojson_newyork <- setNames(geojson_newyork, c("borough_code", "borough", "shape_area", "shape_leng"))
geojson_newyork$borough <- as.factor(geojson_newyork$borough)
geojson_newyork$borough_code <- NULL
head(geojson_newyork)
## borough shape_area shape_leng
## 1 Staten Island 1623620725.05 325917.35395
## 2 Manhattan 636520502.758 357713.308162
## 3 Bronx 1187174772.5 463180.579449
## 4 Brooklyn 1934138215.76 728146.574928
## 5 Queens 3041418506.64 888199.731385
And now we merge geojson_newyork with count_inc_brough maintaining the Spartial Data Frame type.
geojson_newyork@data = data.frame(geojson_newyork@data, count_inc_brough[match(geojson_newyork@data$borough, count_inc_brough$borough),])
geojson_newyork@data$borough.1 <- NULL
And finally we can plot the interactive map using the mapview function.
mapview(list(firefighter_stations_sdf, geojson_newyork),
zcol = list(NULL, "incident_count"),
legend = list(FALSE, TRUE),
homebutton = list(FALSE, TRUE), layer.name = list(NULL, "indicents_number"), alpha.regions = 0.5, aplha = 1)
